Digital mean filter can effectively reduce noise and improve signal-to-noise ratio in high-precision temperature measurement systems, but it will also cause signal distortion and become a source of uncertainty. The collected temperature signal is generally discrete-time series signal, so the existing filter evaluation method is difficult to quantify the distortion caused by this kind of signal. In order to solve this problem, the temperature variation characteristics of the object with slowly varying temperature are analyzed in this work. The typical temperature signal sequence of the measured object is collected by low-noise and high-precision measuring instrument, and then the input sequence of digital mean filter is constructed. A method to evaluate the uncertainty introduced by digital mean filter in high precision temperature measurement system is established. Temperature measurement experiments on a black-body under stable temperature conditions are performed. The temperature range of this black-body is (30.874±0.002)℃, and its rate of the temperature change obeys normal distribution function, which verifies the correctness of the proposed temperature characteristics model. Fluke 1595A thermometer is used to collect typical temperature sequences, and the uncertainty of digital mean filter is analyzed. Under the condition of different sampling intervals and filter lengths, the uncertainty introduced by digital mean filter in high-precision temperature measurement system is difficult to evaluate is basically solved.